Germ line landscape stratification defines distinct molecular and prognostic groups in therapy-related myeloid neoplasms

Key Points

• •Germ line variants stratify t-MN into molecularly and prognostically distinct subgroups beyond treatment-based classifications.
• •CT exposure drives complex karyotypes and poor prognosis in patients with cancer with germ line mutations in cancer-related genes.

Visual Abstract

Abstract

Exposure to chemotherapy (CT) and/or radiotherapy increases the risk of therapy-related myeloid neoplasms (t-MN), a heterogeneous group of disorders currently classified by treatment history rather than molecular features. Although germ line predisposition has been suggested in ∼20% of patients, its prevalence and clinical impact remain insufficiently defined. To address this, we analyzed 100 patients with t-MN by integrating clinical characteristics, prior treatment regimens, and genomic profiling. Somatic and germ line variants were identified using targeted next-generation sequencing (n = 33) or whole-exome sequencing (n = 67). Somatic abnormalities, including cytogenetic and/or molecular alterations, were detected in 89.8% of patients. Germ line variants were identified in 32.3% of patients, including mutations in cancer predisposition genes (19.8%) and myeloid disease–related genes (14.6%). Patient stratification by germ line landscape and gene category defined 2 prognostic scenarios and 3 subgroups with distinct clinical outcomes. Patients harboring germ line cancer predisposition variants were enriched for TP53 mutations and complex karyotypes and had an adverse prognosis. In contrast, patients with germ line myeloid-related variants and those without detectable germ line variants exhibited recurrent mutations in TET2, DNMT3A, SF3B1, SRSF2, RUNX1, and ASXL1, were associated with normal karyotypes, and had favorable outcomes. These findings underscore the importance of the germ line landscape in t-MN pathogenesis and support its incorporation into disease classification and risk stratification. Notably, CT exposure was associated with complex karyotypes and poor prognosis in patients with germ line cancer predisposition variants, highlighting a subgroup that may benefit from targeted surveillance and preventive strategies.

Introduction

Exposure to cytotoxic therapy in patients with a primary disease (PD) confers a future risk of developing the so-called therapy-related myeloid neoplasms (t-MN). t-MN exhibit molecular profiles that differ from those of primary MN, including an enrichment of high-risk cytogenetic alterations and gene mutations.¹ Although t-MN are not considered a single distinct entity in the latest World Health Organization² classification and the International Consensus Classification,³ both emphasize the importance of indicating therapy relatedness in all MN patients arising after prior cytotoxic therapy for an unrelated condition. Recent studies have emphasized that t-MN are not limited to prior neoplastic diseases.4, 5, 6 In addition, less-intensive cytotoxic therapies, such as targeted agents and immunotherapies, have also been associated with an increased risk of t-MN.⁷ The period between exposure and disease onset (latency) varies depending on the type of agent, ranging from 5 to 7 years for alkylating agents and 1 to 3 years for topoisomerase II or poly(ADP-ribose) polymerase inhibitors.⁸

t-MN make up 10% to 20% of newly diagnosed myelodysplastic syndromes (MDS), myelodysplastic/myeloproliferative neoplasms, and acute myeloid leukemia (AML).⁹ Although t-MN classification is not determined by the presence of specific genetic alterations, ∼80% of t-MN present chromosomal abnormalities.¹⁰ Complex karyotypes are detected in 40% of patients with t-MN compared with 5% to 10% of primary MDS,^10,11 which leads to their classification as a high-risk entity with poor prognosis.

t-MN show distinctive mutational signatures compared to primary MN.¹² Concretely, TP53 mutations are markedly enriched in t-MN, likely driven by the selective advantage conferred to preexisting TP53-mutated clones under cytotoxic stress.¹³ These patients exhibit genomic instability and complex karyotypes, which are associated with adverse outcomes.¹⁴ Besides TP53, mutations in PPM1D, ASXL1, and RUNX1 genes are more prevalent in t-MDS than in primary MDS,¹⁵ whereas SF3B1, KRAS, NRAS, and WT1 mutations are more frequent in t-AML than in primary AML.¹⁶

Other factors beyond treatment regimen have been suggested to influence disease development. Indeed, t-MN disease is considered a multifactorial process resulting from complex interactions between underlying germ line susceptibility, stepwise acquisition of somatic mutations, the clonal selection pressure exerted by cytotoxic therapies, alterations of the bone marrow (BM) microenvironment, as well as potentially reduced tolerance to intensive chemotherapy (CT).¹⁵ Notably, between 9% and 31% of t-MN patients harbor germ line mutations,¹⁷ predominantly in genes involved in the DNA damage repair pathway.^10,18

In summary, t-MN classification is mainly based on treatment history, with limited integration of molecular features. In this regard, Sperling et al¹⁹ proposed that genetic profiling may help distinguish true t-MN from secondary sporadic AML or MDS arising coincidentally after cytotoxic exposure. Similarly, we introduced the term “t-MN–like” for patients who develop PD followed by an MN not directly attributable to therapy.²⁰

Given the marked clinical and molecular heterogeneity of t-MN and the underexplored role of germ line mutations and their association with acquired alterations and treatment, improved subclassification is needed. To address this, we performed a comprehensive genomic analysis of 100 patients with t-MN to characterize their cytogenetic and molecular landscape and propose a refined subclassification to better guide treatment strategies for PD in at-risk patients and later disease monitoring.

Materials and methods

Patients and samples

A multicenter cohort of 100 patients with post–cytotoxic therapy MN was analyzed. Clinical data are provided in supplemental Table 1. BM samples were obtained at t-MN diagnosis, and paired peripheral blood (PB) samples were collected at PD and t-MN diagnosis. CD3⁺ T cells from PB were used as a source of germ line DNA. Additional details are provided in the supplemental Methods.

Sequencing studies

Samples were analyzed by targeted next-generation sequencing (tNGS) or whole-exome sequencing (WES) to characterize the molecular landscape (supplemental Figure 1). BM and PB from 33 patients were sequenced by tNGS with a pan-cancer and myeloid gene panels, respectively, whereas WES was performed in 67 patients using paired BM and CD3⁺ T cells. Bioinformatic processing was conducted using in-house pipelines. Variant allele frequency thresholds of ≥2% and ≥5% were applied for tNGS and WES studies.²¹ Additional details are provided in the supplemental Methods.

Molecular evaluation

Two different studies were performed as follows:

• 1.Analysis restricted to variants in genes covered by the 146- and 81-gene targeted panels for both the tNGS and WES series (N = 100; supplemental Tables 2 and 3).
• 2.Analysis of the WES-only cohort (N = 67), considering additional genes reported in the literature or databases with stablished relevance to hematologic or myeloid disorders.

Clusterization and statistics

Acquired alterations were grouped by germ line status and karyotype complexity, alongside clinical profiles, to enable phenotype-genotype analyses. Heat maps were generated in R using the pheatmap package. Clinical features analyses were performed in R statistical platform version 3.6.1. Additional details are provided in the supplemental Methods.

Results

PD and t-MN characteristics

A cohort of 100 patients with t-MN was included in the study (supplemental Table 1). The mean age at onset of the PD was 63.2 years (95% confidence interval [CI], 60.1-66.3). PD treatment included CT in 55 patients (55.0%), radiotherapy (RT) in 15 (15.0%), combined CT + RT in 20 (20.0%), and immunotherapy (IT) in 10 (10.0%) (Table 1). The mean age at t-MN diagnosis was 70.3 years (range, 38-96; 95% CI, 67.5-73.1), with a median latency of 7.3 years (range, 0.4-29). Furthermore, 66 patients (66.6%) with t-MN exhibited an altered karyotype, and among them, 27 patients (40.5%) showed a complex karyotype.

Table 1.

Clinical and treatment characteristics of the patient series according to sequencing method

Series	WES, n	tNGS, n	Total, n	Total, %
No. of patients	67	33	100
Male	32	17	49	49.0
Female	35	16	51	51.0
PD
Age at onset, y	63.21	63.23	63.22
Treatment
CT	28	27	55	55.0
RT	15	0	15	15.0
CT + RT	20	0	20	20.0
IT	4	6	10	10.0
Type
Hematologic	15	1	16	16.0
Lymphoma	12	0	12	12.0
Other	2	0	2	2.0
Multiple∗	1	1	2	2.0
Solid disease	52	32	84	84.0
Breast	17	1	18	18.0
Female reproductive cancers†	6	6	12	12.0
Prostate	11	0	11	11.0
Colorectal‡	4	5	9	9.0
Bladder	2	0	2	2.0
Lung	0	5	5	5.0
Melanoma	0	3	3	3.0
Sarcoma	0	3	3	3.0
Multiple§	8	8	16	16.0
Other	4	1	5	5.0
t-MN
Age at onset, y	72.0	66.9	70.3
Latency, y	9.1	3.6	7.3
Type	67	33	100
MDS-LB-SF3B1	5	3	8	8.0
MDS-BiTP53	5	5	10	10.0
MDS-IB1	6	3	9	9.0
MDS-IB2	7	5	12	12.0
MDS-del(5q)	6	0	6	6.0
CMML	7	1	8	8.0
AML	6	1	7	7.0
MDS-LB	23	14	37	37.0
Other	2	1	3	3.0
Karyotype
Normal	22	11	33	33.0
Altered (1-2 alt)	30	9	39	39.0
Complex (≥3 alt)	15	12	27	27.0
ND	0	1	1	1.0

alt, alterations; CMML, chronic myelomonocytic leukemia; MDS-BiTP53, MDS with biallelic inactivation of TP53; MDS-del(5q), MDS with deletion of 5q arm; MDS-IB1, MDS with increased blasts type 1; MDS-IB2, MDS with increased blasts type 2; MDS-LB, MDS with low blasts; MDS-LB-SF3B1, MDS with low blasts and SF3B1 gene mutation; ND, not documented.

^∗Multiple tumors, including a first treated hematologic tumor.

^†Ovarian, cervical, endometrial, and vulvar cancers were collectively categorized as female reproductive cancers.

^‡Colon and rectal primary tumors were grouped together and classified as colorectal cancer.

^§Multiple tumors, including a first treated solid tumor.

Solid PD predominated across all t-MN subtypes (supplemental Figure 2A). Notably, t-MDS with low blasts and SF3B1 mutations occurred exclusively in patients with a prior solid PD (n = 8), whereas MDS low blast was the most prevalent subtype regardless of PD type (n = 36; supplemental Figure 2B). MDS with isolated 5q deletion showed a strong association with prior exposure to combined CT + RT treatment (n = 5 [83.3%]). In contrast, patients with AML (n = 6 [85.7%]), MDS with increased blasts type 2 (n = 10 [83.3%]), and MDS with biallelic inactivation of TP53 (n = 7 [70.0%]) were predominantly associated with CT-based regimens (Figure 1A).

Figure 1.

Clinical characterization of patients with t-MN. (A) Percentage of patients per treatment regimen across the different t-MN subtype. (B) Distribution of latency periods per patient (in years) across the different PD treatment regimens. (C) Distribution of age at onset (in years) of PD and t-MN across the different PD treatment regimens. (D) Distribution of the number of karyotype alterations per patient according to PD treatment regimen. (E) Relative frequency of specific cytogenetic alterations in those patients with an altered karyotype, stratified per PD treatment regimen. (F) OS analysis based on karyotype complexity. Significant differences were observed between patients, stratified per number of karyotype alterations (P < .0001). (G) OS analysis based on the PD treatment regimen. Altered karyotype indicates 1 to 2 alterations, whereas complex karyotype indicates ≥3 alterations. ∗∗P < .01. CMML, chronic myelomonocytic leukemia; MDS-BiTP53, MDS with biallelic TP53 inactivation; MDS-del(5q), MDS with isolated del(5q); MDS-IB1, MDS with increased blasts 1; MDS-IB2, MDS with increased blasts 2; MDS-LB, MDS with low blasts; MDS-LB-SF3B1, MDS with low blasts and SF3B1 gene mutation.

Latency differed significantly among treatment groups (P = .0004), with CT + RT showing longer latency than CT (P = .0003) and IT (P = .0074; Figure 1B). Age-dependent latency patterns also differed at t-MN onset (P = .027) between the different treatment regimens (Figure 1C). Patients who received CT developed t-MN at a younger age than those treated with RT (P = .023; Tukey honestly significant difference test). A multivariable linear regression analysis assessed factors independently associated with latency after adjustment for age and other covariates (supplemental Figure 3A-B). The model was significant (F test; P < .001) and explained the moderate proportion variability (adjusted R² = 0.27). Increasing age was associated with shorter latency (β = −0.20 per year; P < .001). CT + RT (β = +6.59; P < .001) and RT alone (β = +3.98; P = .021) were associated with longer latencies, whereas IT showed no effect (β = +0.05; P = .98). PD type was not associated with latency (β = −0.93; P = .55; supplemental Figure 3).

Cytogenetic features

CT was associated with an increased number of chromosomal abnormalities (2.7 on average), whereas IT patients showed the lowest number (1.1 on average; Figure 1D). The average number of chromosomal alterations was similar between the RT and CT + RT groups (1.33 and 1.25, respectively). Distinct karyotypic patterns were enriched across treatment groups: −7/del(7q), nonisolated del(5q), and del(13q) were more frequent in CT-treated patients, whereas loss of the Y chromosome was more prevalent among RT-treated patients (Figure 1E). Combined CT + RT regimens were mainly associated with isolated del(5q) and +8 abnormalities.

The median follow-up time for the cohort was 24 months (95% CI, 16-47). Patients with normal karyotypes had significantly improved overall survival (OS; median, 80 months [95% CI, 37 to not reached (NR)]), whereas those with complex karyotypes exhibited the shortest OS (median, 10 months [95% CI, 6-18]; Figure 1F). PD treatment type significantly influenced survival outcomes (P = .014; Figure 1G). The longest OS was observed for RT-treated patients (median, 105 months [95% CI, NR to NR]), whereas CT-treated patients exhibited the poorest survival (median, 17 months [95% CI, 12-32]). CT + RT–treated patients showed relatively favorable outcomes (median, 48 months [95% CI, 39 to NR]), consistent with their low chromosomal alteration burden (Figure 1D). The impact of PD clinical features on t-MN OS adjusted for age was assessed using multivariate Cox regression (supplemental Figure 3C). Age at onset and PD type were not significantly associated with OS, whereas treatment type remained significantly associated. CT + RT (hazard ratio, 0.34; 95% CI, 0.14-0.86; P = .023) and RT alone (hazard ratio, 0.28; 95% CI, 0.08-0.96; P = .042) were associated with reduced mortality risk, consistent with the Kaplan-Meier curves (Figure 1G).

Molecular characterization

A total of 41 pathogenic or likely pathogenic germ line mutations were identified in 31 of 96 evaluable patients (32.3%). In total, 19 mutations were annotated in cancer-related genes (19 patients), including TP53 (n = 7), BRCA2 (n = 5), RAD51C (n = 1), ATR (n = 1), ATM (n = 1), PMS2 (n = 1), CHEK2 (n = 1), MSH2 (n = 1), and BRCA1 (n = 1). The remaining 22 germ line mutations were annotated as myeloid related (14 patients; supplemental Tables 1 and 4). Mutation types per gene are summarized in supplemental Figure 4A. No significant differences were observed in age at PD or t-MN onset, latency periods (Figure 2A), karyotype complexity, or Revised International Prognostic Scoring System (IPSS-R) risk category between patients with and without germ line mutations (supplemental Table 5). Neither age at PD or t-MN onset nor latency differed when patients were stratified by germ line mutations in cancer-related (Figure 2B) or myeloid-related genes (Figure 2C). However, trends suggested earlier PD and t-MN onset in patients with cancer-related gene mutations (60.1 and 67.5 years, respectively) and later onset in those with myeloid-related gene mutations (66.2 and 74.2 years, respectively).

Figure 2.

Molecular characterization of patients with t-MN. (A) Distribution of age at onset (in years) of PD and t-MN in patients with germ line mutations (red) and patients without germ line alterations (black). The bars represent age-dependent latency periods (in years). (B) Distribution of age at onset (in years) of PD and t-MN of patients with germ line mutations in cancer-related genes (red) and patients without them (black). (C) Distribution of age at onset (in years) of PD and t-MN of patients with germ line mutations in myeloid-related genes (red) and patients without them (black). (D) Venn diagram displaying the number of patients harboring germ line mutations in cancer-related genes, myeloid-related genes, and somatic mutations. (E) Distribution of somatic mutations (x-axis) across the genes included in the panel according to germ line scenarios (y-axis), including cancer- or myeloid-related germ line alterations or the absence of germ line variants. The blue gradient indicates the percentage of patients with somatic mutations, ranging from dark blue (enrichment) to light blue (no enrichment). The median number of somatic variants per patient according to germ line mutation status is shown in the SMB column. (F) OS analysis of patients with t-MN stratified by the combination of somatic and germ line mutation status. (G) OS analysis of patients with t-MN stratified according to the presence of germ line mutations in cancer-related or myeloid-related genes compared with patients without germ line mutations.

Regarding the somatic landscape, 123 mutations in myeloid-related genes were identified in 64 of 99 evaluable patients (64.6%), corresponding to a mean of 1.92 mutations per patient with mutations and 1.24 mutations per patient across the cohort (supplemental Table 3). The most frequently mutated genes were TP53 (n = 20), TET2 (n = 14), DNMT3A (n = 10), SF3B1 (n = 9), ASXL1 (n = 8), RUNX1 (n = 8), and SRSF2 (n = 7; supplemental Table 4; supplemental Figure 4B). Overall, when considering both somatic mutations and cytogenetic abnormalities, acquired alterations were detected in 89.8% of patients.

No significant differences in age at t-MN onset were observed between patients with and without somatic mutations (70.2 vs 72.1 years; P = .4; supplemental Figure 5A), even when restricting the analysis to patients without germ line mutations (supplemental Figure 5B).

When stratified by germ line status, somatic mutations were identified in 16.7% and 12.5% of patients with cancer-related and myeloid-related germ line variants, whereas 58.3% had somatic mutations in the absence of germ line alterations (Figure 2D). The mean number of somatic variants per patient was 1.16, 0.85, and 1.21 in these respective groups (Figure 2E). TP53 and PTPN11 mutations were mainly observed in patients with cancer-related germ line variants, although PTPN11 enrichment was driven by 2 mutations in a single case (TRMN_67; supplemental Table 1). TET2 mutations were associated with myeloid-related germ line mutations, whereas DNMT3A mutations were particularly enriched in patients harboring germ line mutations. In contrast, ASXL1, RUNX1, and SF3B1 mutations predominated in patients without germ line mutations (Figure 2E).

Clinical outcome

OS was analyzed across the different germ line and somatic mutation scenarios, with a median follow-up time of 24 months (95% CI, 16-48). Patients without mutations exhibited the most favorable OS (49 months; 95% CI, 39 to NR), followed by those with somatic mutations without germ line mutations (32 months; 95% CI, 18 to NR). Patients with germ line mutations and additional somatic mutations showed poorer prognosis (15 months; 95% CI, 7 to NR; Figure 2F). When stratifying patients according to the category of the germ line–mutated genes, those with germ line mutations in cancer-related genes showed significantly worse OS (10 months; 95% CI, 6 to NR; P = .0063), whereas patients without germ line mutations demonstrated the most favorable outcomes (48 months; 95% CI, 28 to NR; Figure 2G).

Statistical comparison across germ line–stratified subgroups revealed significant differences in karyotype complexity (P = .0061), number of alterations (P = .0038), and frequency of somatic TP53 mutations (P = .0440; Table 2). Post hoc evaluation indicated that these differences were driven by patients with cancer-related germ line mutations, whereas patients with myeloid-related germ line mutations did not differ significantly from those without germ line variants.

Table 2.

Clinical parameters of patients with t-MN per category between subgroups

Subgroups	A: t-MN with germ line mutations in cancer genes	B: t-MN with germ line mutations in myeloid genes	C: t-MN without germ line mutations	P value (all)	P value (A vs C)	P value (B vs C)	P value (A vs B)
Patients, n	18	12	68
Male, %	33.3	66.7	23.5	.0110∗	.3966	.0026∗	.073
Female, %	66.7	33.3	76.5
PD
Age at onset, y	60.1	66.6	63.8	.3546
Treatment, %				.1697
CT	77.8	50.0	50.0
RT	0.0	16.7	19.1
CT + RT	16.7	33.3	17.6
IT	5.6	0.0	13.2
Type, %				.1650
Hematologic disease	11.1	0.0	20.6
Solid disease	88.9	100.0	79.4
t-MN
Age at onset, y	66.5	74.3	71.1	.1759
Latency, y	7.4	7.7	7.2	.9729
MN subtype, %				.5567
MDS-LB-SF3B1	5.6	16.7	7.4
MDS-BiTP53	22.2	0.0	8.8
MDS-IB1	5.6	8.3	10.3
MDS-IB2	16.7	16.7	10.3
MDS-del(5q)	0.0	0.0	8.8
CMML	16.7	16.7	4.4
AML	5.6	8.3	7.4
MDS-LB	27.8	33.3	38.2
Other	0.0	0.0	4.4
Karyotype, %				.0061∗	.0048∗	.3861	.0152∗
Normal	4	6	22
Altered (1-2 alt)	3	5	31
Complex (≥3 alt)	11	1	15
Number of alt	3.7	0.8	1.8	.0038∗	.0039∗	.1725	.0052∗
IPSS-R score, %				.0829
Very high	57.1	11.1	21.0
High	7.1	22.2	12.9
Intermediate	7.1	22.2	16.1
Low	7.1	22.2	37.1
Very low	21.4	22.2	12.9
Patients with mutations, %
Germ line cancer genes	100.0	0.0	0.0	NA
Germ line myeloid genes	11.1	100.0	0.0	NA
Somatic myeloid genes	72.2	75.0	60.3	.4571
Somatic TP53 mutations	33.3	0.0	19.1	.0440∗	.157	.195	.027∗
Somatic mutations per patient	1.3	0.8	1.3	.9059
Somatic mutations per patient with mutations	1.8	1.1	2.1	.3208
Hematologic parameters
BM blasts, %	8	6.8	6.9	.9675
White blood cell count, ×109/L	6.1	6.0	4.5	.1745
Hemoglobin, g/dL	9.6	10.7	9.8	.2322
MCV, fL	91.9	101.9	99.6	.1199
Platelet count, ×109/L	65.0	104.2	135.9	.115
Neutrophils, ×109/L	2.5	3.6	2.6	.5834
Monocytes, ×109/L	1.1	1.0	0.5	.0082∗	.0061∗	.025∗	.4542
OS	Poor (Figure 2G)	Favorable (Figure 2G)	Favorable (Figure 2G)

alt, alterations; CMML, chronic myelomonocytic leukemia; MCV, mean corpuscular volume; MDS-BiTP53, MDS with biallelic inactivation of TP53; MDS-del(5q), MDS with deletion of 5q arm; MDS-IB1, MDS with increased blasts type 1; MDS-IB2, MDS with increased blasts type 2; MDS-LB, MDS with low blasts; MDS-LB-SF3B1, MDS with low blasts and SF3B1 gene mutation; NA, not applicable.

^∗Indicates significant values (P < .05).

Although IPSS-R scores did not reach statistical significance between groups (P = .0829), standardized residuals analysis showed enrichment of the high-risk category (+2.9) and depletion in the low-risk categories (−2.1) with cancer-related germ line mutations. Conversely, depletion of the high-risk category was observed in patients with germ line mutations in myeloid genes (−1.1) and in those without germ line mutations (−1.7). Patients without germ line mutations also showed an enrichment in the low-risk category (+2.1; supplemental Figure 3D). Overall, these findings reveal 2 prognostic t-MN groups: a high-risk group with cancer-related germ line mutations and shorter OS and a lower-risk group with myeloid germ line mutations or no germ line mutations showing favorable outcomes. These differences were not explained by an unequal treatment distribution across groups (P = .1697).

Integrative analysis of clinical outcomes and molecular profiles

The most frequent association was between somatic mutations and either altered or normal karyotypes (Figure 3A). Germ line mutations in myeloid-related genes were mainly associated with normal karyotypes, whereas cancer-related germ line variants were enriched in complex karyotypes. This pattern remained consistent when stratified by PD type, particularly in patients with prior hematologic malignancy (Figure 3B). Patients with altered karyotypes and cancer-related germ line variants showed the lowest somatic mutation burden (SMB; 0.67 variants per patient), followed by those with complex karyotypes (1.00 variants per patient). In contrast, the highest SMB was observed in patients with normal karyotypes regardless of germ line status (1.55 variants per patient; Figure 3C). Gene-specific associations were also evident. In patients with cancer germ line variants, TP53 somatic mutations were enriched in complex karyotypes, whereas PTPN11 were enriched in normal karyotypes. DNMT3A mutations, previously linked to germ line variants (Figure 2E), were enriched in abnormal and complex karyotypes, whereas SF3B1, RUNX1, and ASXL1 were mainly found in patients without germ line mutations and normal karyotypes. TET2 mutations were consistently associated with normal karyotypes.

Figure 3.

Molecular associations. (A) Intersection of patients with germ line (myeloid- or cancer-related genes) and somatic mutations, together with karyotype status (complex, altered [1-2 alterations], or normal). The horizontal bars represent the total number of patients in each mutation or karyotype category, whereas the most frequent combinations of these alterations are identified by dots. The vertical bars indicate the number of patients in each intersecting combination of alterations. (B) Proportion of patients with germ line mutations in myeloid-related and cancer-related genes, stratified by type of PD (solid vs hematologic) and karyotype status. (C) Distribution of somatic mutations across the genes included in the panel (x-axis), stratified by the presence of germ line mutations in cancer- or myeloid-related genes and karyotype complexity (y-axis). The blue gradient indicates the percentage of patients with somatic mutations, ranging from dark blue (enrichment) to light blue (no enrichment). The median number of somatic variants per patient and karyotype complexity is shown in the SMB column. (D) Average SMB in patients with t-MN, stratified by treatment regimen and karyotype complexity. (E) Distribution of somatic mutations across the genes included in the panel (x-axis), stratified by treatment regimen and karyotype complexity (y-axis). The blue gradient indicates the percentage of patients with somatic mutations, ranging from dark blue (enrichment) to light blue (no enrichment). (F) Percentage of CT-treated patients with and without germ line mutations in cancer-related genes, stratified by karyotype status. (G) OS analysis of patients with t-MN with germ line mutations in cancer- or myeloid-related genes compared to those without germ line mutations. Survival differences were assessed using the log-rank test.

SMB according to karyotype complexity was next assessed in relation to treatment regimens (Figure 3D). No significant differences were observed in the overall cohort or within cytotoxic treatment groups (CT, RT, or CT + RT). In contrast, a higher SMB was detected in CT + RT and IT patients, mostly carrying complex and noncomplex karyotypes, respectively. When analyzed at gene level, somatic TP53 and DNMT3A mutations were exclusively enriched in patients receiving CT and with complex and altered karyotypes, respectively. SF3B1 and RUNX1 mutations were more frequent in IT-treated patients with normal karyotypes, whereas ASXL1 mutations were identified in CT-treated patients with normal karyotype (Figure 3E). TET2 mutations remained consistently associated with normal karyotypes across most treatment groups but especially enriched in CT + RT–treated patients. Thus, the most gene-specific associations occurred in CT-treated patients (Figure 3E).

Finally, to assess the role of CT in promoting chromosomal alterations, karyotype complexity was analyzed within the CT subgroup, stratified by germ line mutations (Figure 3F), with patients carrying cancer-related mutations showing higher rates of complex karyotypes. The median follow-up for the CT cohort was 51 months (95% CI, 16 to NR), and OS curves for patients with cancer-related mutations mirrored those of the overall cohort (Figure 2G). In contrast, patients with myeloid-related germ line mutations exhibited superior OS comparable to patients without germ line mutations (Figure 3G).

TP53 in patients with t-MN

Patients with TP53 mutations were compared with those with wild-type (WT) TP53 (WT-TP53). No significant differences were observed in age at PD onset, t-MN onset, or latency (Figure 4A). However, when stratifying according to the germ line or somatic origin, patients carrying germ line TP53 mutations developed PD and t-MN at earlier ages compared to patients with somatic TP53 mutations, although difference did not reach statistical significance (Figure 4B). Multiple patients were more frequently represented in germ line TP53 carriers (28.6%) compared with patients with somatic TP53 mutations (5.8%), potentially reflecting the clinical spectrum of Li-Fraumeni syndrome (supplemental Table 1).

Figure 4.

Characterization of patients with t-MN with TP53 mutations. (A) Distribution of the age at onset (in years) of PD and t-MN in patients with TP53 mutations (red) and patients with WT-TP53 (black). The bars represent age-dependent latency periods (in years). (B) Distribution of the age at onset (in years) of PD and t-MN in patients with germ line TP53 mutations (red) compared to those with somatic TP53 mutations (black). (C) OS analysis of patients with t-MN with TP53 mutations (germ line or somatic) and patients with WT-TP53. Survival differences were assessed using the log-rank test. (D) OS analysis of patients with t-MN with somatic TP53 mutations (stratified by 1-hit or multihit patients) and patients with WT-TP53. Survival differences were assessed using the log-rank test.

OS analysis revealed significantly worse survival for patients with TP53-mutations compared with those with WT-TP53 (48 months; 95% CI, 32 to NR; P < .0001), with similarly poor outcomes among patients with germ line TP53 mutations (7 months; 95% CI, 4 to NR) and somatic TP53 mutations (9 months; 95% CI, 6 to NR; Figure 4C). This effect may reflect the enrichment of germ line cancer-related mutations in patients with somatic TP53 mutations (Figure 2E), with germ line cancer-related mutations themselves being associated with shorter OS (Figure 2F).

Karyotype complexity differed significantly between patients with TP53 mutations and those with WT-TP53, as well as across germ line–somatic stratifications (P < .0001). The proportion of patients with cancer-related germ line comutations also varied among groups (P < .0235), with germ line TP53 carriers showing no additional germ line hits. Patients with somatic TP53 mutations were enriched for cancer-related germ line variants and lower somatic comutations (supplemental Table 6). Although OS did not differ between one-hit and multihit patients (P = .55), both groups had worse outcomes than patients with WT-TP53 (P < .0001; Figure 4D).

WES vs tNGS comparison

WES identified germ line mutations in cancer-related genes in 31.3% of patients, whereas tNGS detected them in 19.8%. Both approaches identified the same proportion of patients with myeloid-related germ line mutations. Notably, WES identified 3 pathogenic germ line mutations in BCHE, 3 in MUTYH, 2 in OGG1, 1 in RTEL1, and 1 in BLM gene. Somatic mutation frequencies were comparable between WES and tNGS (50.7% vs 52.2%), despite WES identifying an additional 1 pathogenic variant in EP300 and 4 in PPM1D genes.

Discussion

Our cohort reflected the heterogeneous clinical outcomes previously described in t-MN, with complex karyotypes and poor prognosis. Breast cancer and non-Hodgkin lymphoma were the most common primary solid and hematologic diseases, respectively, consistent with prior reports.²² Cytotoxic therapy for PD was associated with worse prognosis and increased karyotypic alterations (Figure 1D-F), including −7/del(7q), nonisolated del(5q), and del(13q) (Figure 1E), all linked to poor IPSS-R risk.²³ CT alone conferred the poorest survival (Figure 1G) and earlier t-MN onset compared with RT (P = .023).

Unexpectedly, CT + RT was associated with fewer karyotypic abnormalities, longer latency, and improved OS compared with CT alone. Although CT and RT might be expected to increase cytotoxic synergy,²⁴ combined-modality therapy may reduce cumulative CT exposure, which is the main driver of adverse t-MN outcomes. Prior studies on CT + RT in PD reported improved progression-free survival but did not examine the risk of second malignancies.25, 26, 27 This could reflect that a combined-modality treatment may involve a lower number of CT cycles because CT exposure is the component most clearly linked to adverse prognosis in t-MN.27, 28, 29 Thus, treatment-related factors emerged as the main determinants of latency and outcome in this cohort, whereas age at PD and PD type were not independent predictors of t-MN development or prognosis. Residual confounding from unmeasured variables (eg, performance status) cannot be excluded. Consequently, reducing therapy intensity might be a useful preventive strategy to minimize the risk of t-MN development,²⁸ which is particularly important for high-risk individuals and those carrying germ line mutations in cancer predisposition genes.

Focusing on the molecular landscape, the prevalence of somatic mutations in myeloid-related genes was similar to that previously published for t-MN series,^12,15,29 with a broad mutational spectrum and enrichment of mutations in TP53, TET2, DNMT3A, ASXL1, RUNX1, and SF3B1 (supplemental Figure 6). The proportion of patients with somatic mutations (64.6%) was lower than in most MDS series (80%-90%), likely reflecting methodological differences, because studies lacking paired germ line controls may misclassify germ line as somatic. Accordingly, when all variants are considered regardless of origin, the proportion increases to 72.8%.

TP53 was the most frequently mutated gene in t-MDS (∼40%) compared with de novo MDS (∼10%)³⁰ and was strongly associated with complex karyotypes and low comutations (Figure 3C), supporting a distinct TP53-driven subtype.¹⁴ In addition, both germ line and somatic TP53 mutations were independently associated with poor OS (Figure 4C). Thus, although 1-hit TP53-mutated MDS has previously been reported to have OS comparable to WT-TP53 MDS,³¹ our findings are consistent with recent evidence indicating that 1-hit TP53 mutations are associated with reduced OS in t-MN,^13,32 potentially driven by the high prevalence of complex karyotypes. In addition, mutations in PPM1D gene, a negative regulator of TP53, have been reported more frequently in t-MDS (>10%) than in de novo MDS (∼1%).¹⁶ Although slightly lower, WES identified PPM1D mutations in 3 patients (4.6%), increasing to 6.6% when considering only patients harboring somatic mutations.

Prior studies demonstrated a strong association between cytotoxic treatment, increased karyotypic complexity, and poor prognosis while also describing the molecular landscape of t-MN.^12,30,33 Nevertheless, they have not consistently associated treatment exposures to specific molecular profiles, suggesting that treatment alone does not fully explain t-MN heterogeneity.^15,24 Our study extends these findings by integrating germ line predisposition with clinical outcomes and acquired alterations, highlighting the potential interplay between inherited susceptibility and therapy-induced mutagenesis. Germ line mutations were present in 14.7% of patients in myeloid-related genes, increasing to 33.4% when cancer-related genes were included, suggesting that germ line contribution to t-MN may be underestimated.

Overall clinical and molecular features did not differ between germ line carriers and noncarriers (supplemental Table 5). However, stratification by cancer-related vs myeloid carriers reduced heterogeneity and revealed differences in karyotype complexity, IPSS-R risk, and survival. Germ line cancer-related variants defined a distinct prognostic subgroup characterized by higher IPSS-R scores, frequent somatic TP53 mutations, complex karyotypes, and poor prognosis (Figures 2E-G and 3A-C; Table 2). These findings are consistent with prior evidence that germ line susceptibility increases genomic instability after genotoxic therapy and promotes TP53-driven disease.^5,8,34 Even without a broad spectrum of secondary mutations, TP53 disruption together with cytogenetic instability appears sufficient to drive an aggressive phenotype. Among patients receiving CT, complex karyotypes and adverse outcomes were strongly associated with cancer-related germ line mutations. Overall, inherited predisposition appears to prime hematopoietic progenitors for therapy-related leukemogenesis, favoring evolutionary trajectories with complex karyotypes and poor prognosis.

In contrast, germ line mutations in myeloid genes were associated with normal karyotype and somatic mutations in TET2 and DNMT3A (Figures 2E and 3C), which are frequently observed in MN with germ line RUNX1 or GATA2 deficiencies, SAMD9/SAMD9L-related disorders,³⁵ and familial MDS.³⁶ These observations support that patients with t-MN harboring germ line mutations in myeloid genes may develop MDS through mechanisms comparable to those seen in MDS with germ line predisposition, independent of the PD type or treatment regimen, rendering them clinically and biologically indistinguishable.³⁷ Consistently, the prevalence of germ line mutations in myeloid-related genes in our cohort (14.6%) and the OS were comparable to those in non–germ line t-MN patients (10%-15%).³⁸

The detection of germ line mutations in myeloid genes raises questions about their origin and biological impact because some variants occur at sites typically reported as somatic and are not recognized as germ line predisposition mutations.^2,3 In our cohort, germ line variants detected in CD3⁺ with variant allele frequency <20% accounted for 4.9% of all germ line variants. Given the high CD3⁺ purity (minimum, 85%; mean, 96.9%; 95% CI, 95.35-98.39), crosscontamination is unlikely to fully explain these findings. A more plausible explanation is acquisition in an early hematopoietic progenitor before myeloid-lymphoid divergence, leading to their presence in both compartments.^39,40 Confirmation would require analysis of nonhematopoietic tissues, which were not available.

Despite these considerations, our approach effectively detects previously unrecognized germ line mutations and highlights their impact on both myeloid and lymphoid neoplasms.

Altogether, germ line–based stratification delineates 3 molecular subgroups of t-MN, which fall into 2 prognostic categories (Figure 5).

Figure 5.

Proposed classification of t-MN entity. t-MN with germ line mutations were then stratified, including (1) patients harboring germ line mutations in cancer-related genes, (2) patients harboring germ line mutations in myeloid-related genes, and (3) t-MN without germ line mutations, resembling the patients harboring germ line mutations in myeloid-related genes but driven by cytotoxic exposure. CHIP, clonal hematopoiesis of indeterminate potential.

Germ line mutations in cancer predisposition genes

These patients consistently exhibited complex karyotypes, moderate SMB with enrichment in somatic TP53 mutations, and adverse prognosis. Cancer-predisposing germ line mutations may not only drive PD development but also modulate treatment-related toxicity, ultimately accelerating t-MN onset and promoting cytogenetic instability, with poor outcomes driven by the underlying germ line background rather than by treatment exposure alone.

Germ line mutations in myeloid genes

Although clinically resembling t-MN, these neoplasms arise from germ line myeloid mutations independent of PD type or treatment. These patients were associated with normal karyotypes, low SMB, favorable prognosis, primary solid tumors, and recurrent somatic mutations in TET2 and DNMT3A.

No germ line mutations

In these patients, t-MN onset appeared to be more closely associated with prior cytotoxic treatment exposure because no inherited susceptibility factors were identified. t-MN onset was strictly dependent on cytotoxic treatment. These patients had increased SMB with recurrent mutations in SF3B1, RUNX1, TP53, ASXL1, and SRSF2 genes, normal karyotypes, and favorable prognosis.

Given the relevance of the germ line landscape in explaining heterogeneity, we compared tNGS and WES approaches. Despite both showing comparable sensitivity for somatic variant detection,^22,23 WES provided higher resolution for identifying germ line alterations. Germ line mutations in BCHE and MUTYH have been associated with multiple cancers,^41,42 whereas RTEL1 and OGG1 are involved in telomere maintenance and DNA repair,⁴³ and BLM mutations are associated with increased cancer risk.⁴⁴ These findings support the use of WES in future studies to refine t-MN classification.

Overall, our data underscore the interplay between germ line susceptibility and t-MN development and support the existence of biologically distinct yet clinically homogeneous subgroups with different prognoses. Cytotoxic therapy may therefore play a secondary role in disease initiation and progression. Consistently, germ line–based stratification reduced apparent heterogeneity, with patients lacking cancer-related germ line mutations showing more favorable outcomes than previously reported.

These results challenge the traditional classification of t-MN based solely on prior cytotoxic exposure and emphasize the germ line landscape as a key determinant of disease biology. Accordingly, integrating germ line testing strategies that specifically include cancer predisposition genes, which may not be included in some myeloid next-generation sequencing panels, may improve risk stratification, particularly among patients undergoing cytotoxic therapies and those with t-MN. Furthermore, the analysis of germ line tissue samples may reveal the germ line origin of myeloid-related mutations, including TP53 variants, which would otherwise be assumed to be somatic in routine clinical practice. Moreover, although CT has been associated with increased karyotypic complexity and poorer outcomes, CT + RT showed more favorable karyotypic profiles, latency, and survival, suggesting a potential mitigation of CT-related effects. These observations support therapy deescalation as a potential preventive strategy, particularly for genetically high-risk individuals.

We acknowledge that subgroup analyses were limited by the small cohort size, reducing statistical power. Therefore, these findings should be interpreted cautiously and considered exploratory, pending validation in larger independent cohort, to better define the role of specific genetic alterations across t-MN subgroups.

Conflict-of-interest disclosure: The authors declare no competing financial interests.